Predicting Urban Sprawl Dynamics in Algiers Metropolitan Area Using Multi-Temporal Remote Sensing and Machine Learning: A Land Use Change Modeling Approach

Introduction:

Algiers, the capital of Algeria, has experienced unprecedented urban expansion over the past three decades, driven by demographic pressure and unplanned development. Between 1990 and 2020, the built-up area of Greater Algiers expanded by over 180%, significantly encroaching on fertile agricultural land in the Mitidja plain and vital coastal ecosystems. This study addresses the urgent need for predictive modeling by developing a machine learning-based framework to forecast urban land-use changes through 2040.

Methods:

Multi-temporal Landsat imagery (1990, 2000, 2010, 2020) was processed via Google Earth Engine using a Random Forest classification algorithm to generate high-accuracy land-use maps. Transition probability matrices were derived using Cellular Automata–Markov Chain (CA-Markov) modeling, calibrated with topographic and demographic variables. Furthermore, a Convolutional Neural Network (CNN) was trained to simulate complex urban growth scenarios, including business-as-usual and high-density infill planning policies.

Results:

The Random Forest classifier achieved a robust Overall Accuracy of 94.3% and a Kappa coefficient of 0.91. Simulations indicate a potential 42% increase in built-up areas by 2040 under current trends. Notably, the CNN-based model demonstrated superior spatial accuracy (FoM = 0.38) compared to conventional CA-Markov alone (FoM = 0.24). Scenario analysis reveals that high-density infill policies could reduce peripheral land consumption by up to 61%.

Conclusions:

This study demonstrates the effectiveness of integrating deep learning with geospatial analysis for predicting urban sprawl. The findings highlight the urgent need for compact city strategies to protect the Mitidja agricultural plain and coastal ecosystems. This methodology provides a scalable tool for evidence-based urban planning across metropolitan contexts in the Global South.